Research On Verification And Performance Improvement Methods Of Damage Diagnosis Model For Cable-Stayed Bridges

Damage diagnosis model plays an important role in the health monitoring system of cable-stayed bridge.It is necessary to establish an accurate damage diagnosis model of cable-stayed bridge,objectively evaluate the effectiveness of the model and improve its prediction performance.In this thesis,the data mining method is used to establish the damage diagnosis model of cable-stayed bridge.Taking the two benchmark finite element models of cable-stayed bridge created by Abaqus as the research object,the two methods of damage diagnosis model verification and performance improvement of cable-stayed bridge are studied respectively.The main research work is as follows:(1)The establishment method of cable-stayed bridge damage diagnosis model based on data mining is studied.The internal mechanism of data mining algorithms such as support vector machine,naive Bayes,logistic regression and K-nearest neighbor and their application in cable-stayed bridge damage diagnosis are studied.Python programming language is used to write and debug the code of the algorithm model.According to the unique process of data mining algorithm and the characteristics of cable-stayed bridge,the establishment method of damage diagnosis model of cable-stayed bridge is condensed.(2)The root mean square error(RMSE)evaluation index and the statistical hypothesis test(SHT)are combined to obtain the model validation method(RMSE-SHT,R-S),which is applied to the reliability evaluation of the data mining damage diagnosis model of cable-stayed bridge.Firstly,the R-S method evaluates the diagnostic effect of the model by the evaluation index.Secondly,the SHT method is used to quantitatively compare the consistency between the output data of the diagnostic model and the actual output data of the structure.Finally,the validity of the model is evaluated quantitatively.An example is used to demonstrate the application of the proposed method.The results show that the diagnostic model has high reliability and the confidence level P_δis over 90%,which indicates the effectiveness of the diagnostic model verification method.(3)Based on the improved Kmeans SMOTE method,a new method(VF-Kmeans SMOTE,V-Kmeans SMOTE)is proposed to improve the performance of diagnostic models by combining variance filtering(VF)feature selection technology.This method filters out the zero-variance features in displacement,acceleration and other data,and oversamples the damage data samples corresponding to the filtered features.The method is applied to an H-shaped cable-stayed bridge as a numerical example.The results show that after applying the V-Kmeans SMOTE method,the classification accuracy,precision and F1 score of the support vector machine classification model are increased by 6.19%,7.93%and 20.07%on average on the data set,and the K-nearest neighbour model is increased by 6.18%,7.23%and 7.26%on average,respectively,which verifies the effectiveness of the proposed method.